INDEX

Explanations

words related to provocative and controversial topics or themes

oai_token-act-pair · gpt-3.5-turbo

Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.8.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Negative Logits

glas

-0.80

aird

-0.75

frames

-0.72

ĸļ

-0.69

 Prospect

-0.67

ITNESS

-0.65

 Backup

-0.64

 cushion

-0.64

RFC

-0.63

ynthesis

-0.63

POSITIVE LOGITS

 blasp

1.08

 perpetrated

1.05

 slander

0.99

 punishable

0.94

 slurs

0.93

 inflicted

0.93

ophobic

0.92

uous

0.91

ocaust

0.90

 hatred

0.88

Activations Density 0.171%

words related to provocative and controversial topics or themes

No Known Activations